The 400d Galaxy Cluster Survey weak lensing programme

Durham Research Online
Deposited in DRO:
05 August 2014
Version of attached le:
Published Version
Peer-review status of attached le:
Peer-reviewed
Citation for published item:
Israel, H. and Reiprich, T.H. and Erben, T. and Massey, R.J. and Sarazin, C.L. and Schneider, P. and
Vikhlinin, A. (2014) 'The 400d Galaxy Cluster Survey weak lensing programme. III. Evidence for consistent
WL and X-ray masses at z
0.5.', Astronomy and astrophysics., 564 . A129.
Further information on publisher's website:
http://dx.doi.org/10.1051/0004-6361/201322870
Publisher's copyright statement:
Reproduced with permission from Astronomy
Astrophysics,
c
ESO
Additional information:
Use policy
The full-text may be used and/or reproduced, and given to third parties in any format or medium, without prior permission or charge, for
personal research or study, educational, or not-for-prot purposes provided that:
• a full bibliographic reference is made to the original source
• a link is made to the metadata record in DRO
• the full-text is not changed in any way
The full-text must not be sold in any format or medium without the formal permission of the copyright holders.
Please consult the full DRO policy for further details.
Durham University Library, Stockton Road, Durham DH1 3LY, United Kingdom
Tel : +44 (0)191 334 3042 | Fax : +44 (0)191 334 2971
http://dro.dur.ac.uk
Astronomy
&
Astrophysics
A&A 564, A129 (2014)
DOI: 10.1051/0004-6361/201322870
c ESO 2014
The 400d Galaxy Cluster Survey weak lensing programme
III. Evidence for consistent WL and X-ray masses at z ≈ 0.5?
Holger Israel1,2 , Thomas H. Reiprich2 , Thomas Erben2 , Richard J. Massey1 , Craig L. Sarazin3 ,
Peter Schneider2 , and Alexey Vikhlinin4
1
2
3
4
Department of Physics, Durham University, South Road, Durham DH1 3LE, UK
e-mail: [email protected]
Argelander-Institut für Astronomie, Auf dem Hügel 71, 53121 Bonn, Germany
Department of Astronomy, University of Virginia, 530 McCormick Road, Charlottesville VA 22904, USA
Harvard-Smithsonian Center for Astrophysics, 60 Garden Street, Cambridge MA 02138, USA
Received 17 October 2013 / Accepted 11 February 2014
ABSTRACT
Context. Scaling properties of galaxy cluster observables with cluster mass provide central insights into the processes shaping
clusters. Calibrating proxies for cluster mass that are relatively cheap to observe will moreover be crucial to harvest the cosmological
information available from the number and growth of clusters with upcoming surveys like eROSITA and Euclid. The recent Planck
results led to suggestions that X-ray masses might be biased low by ∼40%, more than previously considered.
Aims. We aim to extend knowledge of the weak lensing – X-ray mass scaling towards lower masses (as low as 1 × 1014 M ) in
a sample representative of the z ∼ 0.4–0.5 population. Thus, we extend the direct calibration of cluster mass estimates to higher
redshifts.
Methods. We investigate the scaling behaviour of MMT/Megacam weak lensing (WL) masses for 8 clusters at 0.39 ≤ z ≤ 0.80 as
part of the 400d WL programme with hydrostatic Chandra X-ray masses as well as those based on the proxies, e.g. YX = T X Mgas .
Results. Overall, we find good agreement between WL and X-ray masses, with different mass bias estimators all consistent with zero.
When subdividing the sample into a low-mass and a high-mass subsample, we find the high-mass subsample to show no significant
mass bias while for the low-mass subsample, there is a bias towards overestimated X-ray masses at the ∼2σ level for some mass
proxies. The overall scatter in the mass-mass scaling relations is surprisingly low. Investigating possible causes, we find that neither
the greater range in WL than in X-ray masses nor the small scatter can be traced back to the parameter settings in the WL analysis.
Conclusions. We do not find evidence for a strong (∼40%) underestimate in the X-ray masses, as suggested to reconcile recent
Planck cluster counts and cosmological constraints. For high-mass clusters, our measurements are consistent with other studies in the
literature. The mass dependent bias, significant at ∼2σ, may hint at a physically different cluster population (less relaxed clusters with
more substructure and mergers); or it may be due to small number statistics. Further studies of low-mass high-z lensing clusters will
elucidate their mass scaling behaviour.
Key words. galaxies: clusters: general – cosmology: observations – gravitational lensing: weak – X-rays: galaxies: clusters
1. Introduction
Galaxy cluster masses hold a crucial role in cosmology. In the
paradigm of hierarchical structure formation from tiny fluctuations in the highly homogeneous early cosmos after inflation,
clusters emerge via the continuous matter accretion onto local
minima of the gravitational potential. Depending sensitively on
cosmological parameters, the cluster mass function, i.e. their
abundance as function of mass and redshift z, provides observational constraints to cosmology (e.g., Vikhlinin et al. 2009b;
Allen et al. 2011; Planck Collaboration XX 2014).
Observers use several avenues to determine cluster masses:
properties of the X-ray-emitting intracluster medium (ICM), its
imprint on the cosmic microwave background via the SunyaevZel’dovich (SZ) effect in the sub-mm regime, galaxy richness estimates and dynamical masses via optical imaging and
?
Appendix A is available in electronic form at
http://www.aanda.org
spectroscopy, and gravitational lensing. Across all wavelengths,
cluster cosmology surveys are under preparation, aiming at
a complete cluster census out to ever higher redshifts, e.g.
eROSITA (Predehl et al. 2010; Merloni et al. 2012; Pillepich
et al. 2012) and Athena (Nandra et al. 2013; Pointecouteau
et al. 2013) in X-rays, Euclid (Laureijs et al. 2011; Amendola
et al. 2013), DES and LSST in the optical/near-infrared, CCAT
(Woody et al. 2012) and SKA at sub-mm and radio frequencies.
Careful X-ray studies of clusters at low and intermediate
redshifts yield highly precise cluster masses, but assume hydrostatic equilibrium, and in most cases spherical symmetry (e.g.
Croston et al. 2008; Ettori et al. 2013). Observational evidence
and numerical modelling challenge these assumptions for all but
the most relaxed systems (e.g. Mahdavi et al. 2008; Rasia et al.
2012; Limousin et al. 2013; Newman et al. 2013). While simulations find X-ray masses to only slightly underestimate the true
mass of clusters that exhibit no indications of recent mergers
and can be considered virialised, non-thermal pressure support
Article published by EDP Sciences
A129, page 1 of 15
A&A 564, A129 (2014)
can lead to a >20% bias in unrelaxed clusters (Laganá et al.
2010; Rasia et al. 2012). Shi & Komatsu (2014) modelled the
pressure due to ICM turbulence analytically and found a ∼10%
underestimate of cluster masses compared to the hydrostatic
case.
Weak lensing (WL), in contrast, is subject to larger stochastic uncertainties, but can in principle yield unbiased masses, as
no equilibrium assumptions are required. Details of the mass
modelling however can introduce biases, in particular concerning projection effects, the source redshift distribution and the
departures from an axisymmetric mass profile (Corless & King
2009; Becker & Kravtsov 2011; Bahé et al. 2012; Hoekstra et al.
2013). For individual clusters, stochastic uncertainties dominate
the budget; however, larger cluster samples benefit from improved corrections for lensing systematics, driven by cosmic
shear projects (e.g. Massey et al. 2013).
Most of the leverage on cosmology and structure formation
from future cluster surveys will be due to clusters at higher z
than have been previously investigated. Hence, the average cluster masses and signal/noise ratios for all observables are going
to be smaller. Even and especially for the deepest surveys, most
objects will lie close to the detection limit. Thus the scaling of
inexpensive proxies (e.g. X-ray luminosity LX ) with total mass
needs to be calibrated against representative cluster samples at
low and high z. Weak lensing and SZ mass estimates are both
good candidates as they exhibit independent systematics from
X-rays and a weaker z-dependence in their signal/noise ratios.
Theoretically, cluster scaling relations arise from their description as self-similar objects forming through gravitational
collapse (Kaiser 1986), and deviations from simple scaling laws
provide crucial insights into cluster physics. For the current state
of scaling relation science, we point to the recent review by
Giodini et al. (2013). As we are interested in the cluster population to be seen by upcoming surveys, we focus here on results
obtained at high redshifts.
Self-similar modelling includes evolution of the scaling
relation normalisations with the Hubble expansion, which is
routinely measured (e.g. Reichert et al. 2011; Ettori 2013).
Evolution effects beyond self-similarity, e.g. due to declining AGN feedback at low z, have been claimed and discussed
(e.g. Pacaud et al. 2007; Short et al. 2010; Stanek et al. 2010;
Maughan et al. 2012), but current observations are insufficient
to constrain possible evolution in slopes (Giodini et al. 2013).
Evidence for different scaling behaviour in groups and low-mass
clusters was found by, e.g., Eckmiller et al. (2011), Stott et al.
(2012), Bharadwaj et al. (2014).
Reichert et al. (2011) and Maughan et al. (2012) investigated X-ray scaling relationships including clusters at z > 1,
and both stressed the increasing influence of selection effects
at higher z. Larger weak lensing samples of distant clusters
are just in the process of being compiled (Jee et al. 2011;
Foëx et al. 2012; Hoekstra et al. 2011a, 2012; Israel et al.
2012; von der Linden et al. 2014a; Postman et al. 2012). Thus
most WL scaling studies are currently limited to z . 0.6, and
also include nearby clusters (e.g. Hoekstra et al. 2012; Mahdavi
et al. 2013, M13). The latter authors find projected WL masses
follow the expected correlation with the SZ signal YSZ , corroborating similar results for more local clusters by Marrone
et al. (2009, 2012). Miyatake et al. (2013) performed a detailed
WL analysis of a z = 0.81 cluster discovered in the SZ using
the Atacama Cosmology Telescope, and compared the resulting
lensing mass against the Reese et al. (2012) YSZ –M scaling relation, in what they describe as a first step towards a high-z SZ-WL
scaling study.
A129, page 2 of 15
By compiling Hubble Space Telescope data for 27 massive
clusters at 0.83 < z < 1.46, Jee et al. (2011) not only derive
the relation between WL masses M wl and ICM temperature T X ,
but also notice a good correspondence between WL and hydrostatic X-ray masses M hyd . As they focus on directly testing cosmology with the most massive clusters, these authors however
stop short of deriving the WL-X-ray scaling. Also using HST
observations, Hoekstra et al. (2011a) investigated the WL mass
scaling of the optical cluster richness (i.e. galaxy counts) and LX
of 25 moderate-LX clusters at 0.3 < z < 0.6, thus initiating the
study of WL scaling relations off the top of the mass function.
Comparisons between weak lensing and X-ray masses for
larger cluster samples were pioneered by Mahdavi et al. (2008)
and Zhang et al. (2008), collecting evidence for the ratio of weak
lensing to X-ray masses M wl /M hyd > 1, indicating non-thermal
pressure. Zhang et al. (2010), analysing 12 clusters from the
Local Cluster Substructure Survey (LoCuSS), find this ratio to
depend on the radius. Likewise, a difference between relaxed and
unrelaxed clusters is found (Zhang et al. 2010; Mahdavi et al.
2013). Rasia et al. (2012) show that the gap between X-ray and
lensing masses is more pronounced in simulations than in observations, pointing to either an underestimate of the true mass
also by WL masses (cf. Bahé et al. 2012) or to simulations
overestimating the X-ray mass bias.
The current disagreement between the cosmological constraints derived from Planck primary cosmic microwave background (CMB) data with Wilkinson Microwave
Anisotropy Probe data, supernova data, and cluster data (Planck
Collaboration XX 2014) may well be alleviated by, e.g. sliding
up a bit along the Planck degeneracy curve between the Hubble
factor H0 and the matter density parameter Ωm . Nevertheless,
as stronger cluster mass biases than currently favoured (∼40%)
have also been invoked as a possible explanation, it is very
important to test the cluster mass calibration with independent
methods out to high z, as we do in this work.
This article aims to test the agreement of the weak lensing
and X-ray masses measured by Israel et al. (2012) for 8 relatively low-mass clusters at z & 0.4 with scaling relations from
the recent literature. The 400d X-ray sample from which our
clusters are drawn has been constructed to contain typical objects at intermediate redshifts, similar in mass and redshift to
upcoming surveys. Hence, it does not include extremely massive low-z clusters. We describe the observations and WL and
X-ray mass measurements for the 8 clusters in Sect. 2, before
presenting the central scaling relations in Sect. 3. Possible explanations for the steep slopes our scaling relations exhibit are
discussed in Sect. 4, and we compare to literature results in
Sect. 5, leading to the conclusions
in Sect. 6. Throughout this
p
article, E(z) = H(z)/H0 = Ωm (z + 1)3 + ΩΛ denotes the selfsimilar evolution factor (Hubble factor H(z) normalised to its
present-day value of H0 = 72 km s−1 Mpc−1 ), computed for a
flat universe with matter and dark energy densities of Ωm = 0.3
and ΩΛ = 0.7 in units of the critical density.
2. Observations and data analysis
2.1. The 400d weak lensing survey
This article builds on the weak lensing analysis for 8 clusters of galaxies (Israel et al. 2010, 2012, Paper I and Paper II
hereafter) selected from the 400d X-ray selected sample of
clusters (Burenin et al. 2007; Vikhlinin et al. 2009a, V09a).
From the ∼400 deg2 of all suitable ROSAT PSPC observations,
H. Israel et al.: The 400d Galaxy Cluster Survey weak lensing programme. III.
Burenin et al. (2007) compiled a catalogue of serendipitously
detected clusters, i.e. discarding the intentional targets of the
ROSAT pointings. For a uniquely complete subsample of 36
X-ray luminous (LX & 1044 erg/s) high-redshift (0.35 ≤ z ≤
0.89) sources, V09a obtained deep Chandra data, weighing
the clusters using three different mass proxies (Sect. 3.2).
Starting from the cluster mass function computed by V09a,
Vikhlinin et al. (2009b) went on to constrain cosmological
parameters. For brevity, we will refer to the V09a high-z sample as the 400d sample. The 400d weak lensing survey follows up these clusters in weak lensing, determining independent
WL masses with the ultimate goals of deriving the lensingbased mass function for the complete sample and to perform
detailed consistency checks. Currently, we have determined WL
masses for 8 clusters observed in four dedicated MMT/Megacam
runs (see Papers I and II). Thus, our scaling relation studies are largely limited to this subset of clusters, covering the
sky between αJ2000 = 13h 30m –24h with δJ2000 > 10◦ and
αJ2000 = 0h –08h 30m with δJ2000 > 0◦ .
Bhattacharya et al. (2013, B13)1 . Finding the masses using B01
or B13 makes them less susceptible to variations in the model in
Paper II, we explore their effect further in Sect. 4.1.
2.2. Weak lensing analysis
c∆ r/r∆
4π 3 ln (1 + c∆ r/r∆ ) − 1 + c∆ r/r∆
MNFW (r) = ∆ρc r∆ ×
3
ln (1 + c∆ ) − c∆ /(1 + c∆ )
= M∆ (r∆ ) × Ξ(r; r∆ , c∆ ),
We present only a brief description of the WL analysis in this
paper; for more details see Paper II. Basic data reduction is performed using the THELI pipeline for multi-chip cameras (Erben
et al. 2005; Schirmer 2013), adapted to MMT/Megacam. We employ the photometric calibration by Hildebrandt et al. (2006).
Following Dietrich et al. (2007), regions of the THELI coadded
images not suitable for WL shear measurements are masked.
Shear is measured using an implementation of the “KSB+” algorithm (Kaiser et al. 1995; Erben et al. 2001), the “TS” pipeline
(Heymans et al. 2006; Schrabback et al. 2007; Hartlap et al.
2009). Catalogues of lensed background galaxies are selected
based on the available colour information. For clusters covered in three filters, we include galaxies based on their position in colour–colour–magnitude space (Paper II; see Klein et al.,
in prep., for a generalisation). For clusters covered only in one
passband, we apply a magnitude cut. Where available, colour information also enables us to quantify and correct for the dilution
by residual cluster members (Hoekstra 2007) in the shear catalogues. The mass normalisation of the WL signal is set by the
mean lensing depth hβi, defined as β = Dds /Ds , the ratio of angular diameter distances between the deflector and the source,
and between the observer and the source. The Ilbert et al. (2006)
CFHTLS Deep fields photometric redshift catalogue serves as a
proxy for estimating hβi and for calibrating the background selection.
The tangential ellipticity profiles given the ROSAT cluster centres are modelled by fitting the reduced shear profile
(Bartelmann 1996; Wright & Brainerd 2000) corresponding
to the Navarro et al. (1996, 1997, NFW) density profile between 0.2 Mpc and 5.0 Mpc projected radius. Input ellipticities
are scaled according to the Hartlap et al. (2009) calibration factor
and, where applicable, with the correction for dilution by cluster members. We consider the intrinsic source ellipticity measured from the data, accounting for its dependence on the shear
(Schneider et al. 2000).
Lensing masses are inferred by evaluating a χ2
merit function on a grid in radius r200 and concentration c200 . The latter is poorly constrained in the direct fit,
so we marginalise over it assuming an empirical massconcentration relation. In addition to the direct fit approach,
in Israel et al. (2012), we report masses using two different
mass-concentration relations: Bullock et al. (2001, B01), and
2.3. Choice of the overdensity contrast
Cluster scaling relations are usually given for the mass contained
within a radius r500 , corresponding to an overdensity ∆ = 500
compared to the critical density ρc of the Universe at the cluster redshift. This ∆ is chosen because the best-constrained X-ray
masses are found close to r500 , determined by the particle backgrounds of Chandra and XMM-Newton (cf. Okabe et al. 2010a).
Currently, only Suzaku allows direct constraints upon X-ray
masses at r200 (see Reiprich et al. 2013, and references therein).
In order to compare to the results from the Vikhlinin et al.
(2009a) Chandra analysis, we compute our ∆ = 500 WL masses
from our ∆ = 200 masses, assuming the fitted NFW profiles
given by (r200 , c200 ) to be correct. Independent of ∆ > 1, the cumulative mass of a NFW halo, described by r∆ and c∆ , out to a
test radius r is given by:
(1)
(2)
separating into the mass M∆ and a function we call Ξ(r; r∆ , c∆ ).
Equating Eq. (1) with r = r500 for ∆ = 200 and ∆0 = 500,
we arrive at this implicit equation for r500 , which we solve
numerically:
1/3
r500 = r200 52 Ξ(r500 , r200 , c200 )
.
(3)
2.4. X-ray analysis
Under the strong assumptions that the ICM is in hydrostatic
equilibrium and follows a spherically symmetric mass distribution, the cluster mass within a radius r can be calculated as (see
e.g. Sarazin 1988):
!
−kB T X (r) r d ln ρg (r) d ln T X (r)
hyd
M (r) =
+
(4)
µmpG
d ln r
d ln r
from the ICM density and temperature profiles ρg (r) and T X (r),
where G is the gravitational constant, mp is the proton mass,
and µ = 0.5954 the mean molecular mass of the ICM. The
ICM density is modelled by fitting the observed Chandra surface
brightness profile, assuming a primordial He abundance and an
ICM metallicity of 0.2 Z , such that ρg (r) = 1.274 mp n(r). We
use
p a Vikhlinin et al. (2006) particle density profile with n(r) =
np (r) ne (r). Extending the widely-used β-profile (Cavaliere &
Fusco-Femiano 1978), it allows for prominent cluster cores as
well as steeper surface brightness profiles in the cluster outskirts
to be modelled by additional terms. Regarding the systematic
differences Rozo et al. (2014b) find between X-ray mass algorithms, we point out that we employ the V09a profiles directly
rather than re-deriving them from their parameters.
The relatively low signal/noise in the Chandra data renders
the determination of individual temperature profiles difficult.
Rather, we fit a global T X (Table 1; V09a) and assume the empirical average temperature profile T X (r) = T X (1.19−0.84r/r200 )
Reiprich et al. (2013) derive from compiling all available Suzaku
1
Actually, we use the slightly different relation as given in Version 1,
referred to in Paper II as “B12”: arxiv.org/abs/1112.5479v1
A129, page 3 of 15
A&A 564, A129 (2014)
Table 1. Measured properties of the 400d MMT cluster sample.
CL 0030
+2618
CL 0159
+0030
CL 0230
+1836
CL 0809
+2811
CL 1357
+6232
CL 1416
+4446
CL 1641
+4001
CL 1701
+6414
0.50
5.63 ± 1.13
3.43 ± 0.41
4.41 ± 1.33
2.04 ± 0.19
0.39
4.25 ± 0.96
2.51 ± 0.37
2.67 ± 0.90
1.92 ± 0.22
0.80
5.50 ± 1.02
3.46 ± 0.46
3.57 ± 0.99
2.70 ± 0.27
0.40
4.17 ± 0.73
3.69 ± 0.42
2.96 ± 0.78
3.98 ± 0.35
0.53
4.60 ± 0.69
2.96 ± 0.29
2.78 ± 0.62
2.40 ± 0.18
0.40
3.26 ± 0.46
2.52 ± 0.24
1.76 ± 0.37
3.10 ± 0.24
0.46
3.31 ± 0.62
1.70 ± 0.20
1.73 ± 0.49
1.34 ± 0.13
0.45
4.36 ± 0.46
3.28 ± 0.24
2.66 ± 0.42
3.20 ± 0.20
wl
r200
(cfit ) [kpc] from Paper II
wl
r200
(cB13 ) [kpc] from Paper II
wl
r200
(cfit ), no dilu. corr. [kpc]
wl
r200 (cB13 ), no dilu. corr. [kpc]
1520+140
−160
1445+132
−151
1430+140
−150
1349+123
−136
1370+180
−220
1312+222
−275
1320+170
−200
1265+214
−256
1540+280
−320
1514+273
−317
1470+280
−310
1432+258
−305
1750+230
−280
1714+192
−221
1660+220
−270
1629+182
−204
1110+210
−250
1057+202
−236
–
–
980+150
−180
977+155
−183
–
–
1060+300
−260
1014+191
−126
–
–
940+320
−290
1014+166
−201
–
–
wl
r500
(cfit ) [kpc]
wl
r500
(cB13 ) [kpc]
wl
r500
(cfit ), no dilu. corr. [kpc]
wl
r500
(cB13 ), no dilu. corr. [kpc]
Y
r500
[kpc]
T
r500 [kpc]
G
[kpc]
r500
914+84
−96
933+84
−96
850+83
−89
870+87
−90
873 ± 85
949 ± 72
734 ± 120
959+126
−154
848+142
−175
924+119
−140
816+136
−168
821 ± 109
838 ± 105
751 ± 130
974+177
−203
959+171
−203
919+175
−194
909+165
−191
777 ± 75
785 ± 73
715 ± 88
994+131
−159
1108+129
−142
923+122
−150
1051+116
−129
930 ± 84
864 ± 97
954 ± 80
674+127
−152
685+129
−155
–
–
821 ± 92
804 ± 96
766 ± 106
651+100
−120
637+104
−117
–
–
819 ± 108
727 ± 138
877 ± 94
440+125
−108
655+136
−182
–
–
702 ± 138
706 ± 136
648 ± 161
404+137
−125
655+110
−130
–
–
877 ± 89
818 ± 102
870 ± 91
wl
wl
M500
(r500
), using cfit
wl
wl
M500 (r500
), using cB13
wl
wl
M500
(r500
), using cfit , no dilu. corr.
wl
wl
M500 (r500 ), using cB13 , no dilu. corr.
wl
Y
M500
(r500
), using cB13
wl
T
M500 (r500
), using cB13
wl
G
M500
(r500
), using cB13
3.94+1.19
−1.12
4.19+1.23
−1.16
3.17+1.02
−0.89
3.40+0.98
−0.95
3.91+0.78
−0.83
4.26+0.83
−0.91
3.26+0.69
−0.70
4.00+1.79
−1.63
2.77+1.64
−1.38
3.58+1.57
−1.39
2.46+1.45
−1.23
2.68+0.93
−1.12
2.73+0.95
−1.15
2.45+0.85
−1.02
6.82+4.43
−3.44
6.51+4.14
−3.32
+3.93
5.73−2.92
+3.60
5.54−2.81
5.15+1.83
−2.07
5.21+1.85
−2.10
4.68+1.66
−1.86
4.51+2.03
−1.84
6.24+2.44
−2.11
3.61+1.63
−1.49
+1.97
5.33−1.73
+1.25
5.15−1.34
4.76+1.17
−1.24
5.29+1.28
−1.38
1.64+1.11
−0.88
1.72+1.16
−0.92
–
–
2.04+0.81
−0.96
+0.79
2.00−0.94
+0.75
1.91−0.89
1.27+0.68
−0.58
1.19+0.68
−0.54
–
–
1.49+0.52
−0.59
+0.47
1.34−0.52
1.58+0.56
−0.63
+0.47
0.42−0.24
1.38+1.05
−0.86
–
–
1.48+0.64
−0.85
1.48+0.64
−0.85
1.37+0.59
−0.78
0.32+0.45
−0.22
1.37+0.81
−0.66
–
–
1.78+0.64
−0.76
1.68+0.60
−0.71
1.77+0.64
−0.75
hyd wl
M500
(r500 ), using cfit
hyd wl
(r500 ), using cB13
M500
hyd wl
(r500 ), using cfit , no dilu. corr.
M500
hyd wl
(r500 ), using cB13 , no dilu. corr.
M500
3.46+1.10
−1.01
3.47+1.10
−1.00
3.21+1.05
−0.94
3.22+1.02
−0.94
2.78+1.11
−1.00
2.50+1.12
−1.01
2.68+1.06
−0.94
2.40+1.08
−0.97
5.13+2.95
−2.37
4.99+2.84
−2.33
+2.81
4.66−2.18
+2.64
4.54−2.12
3.66+1.36
−1.26
3.99+1.35
−1.22
3.35+1.27
−1.18
3.75+1.26
−1.13
2.20+0.92
−0.85
2.20+0.91
−0.85
–
–
1.31+0.51
−0.44
1.28+0.51
−0.43
–
–
+0.70
1.17−0.51
1.65+0.75
−0.72
–
–
1.02+0.60
−0.45
1.70+0.53
−0.52
–
–
Redshift z
kB T X from V09a [keV]
Y
M500
from V09a
T
M500 from V09a
G
M500
from V09a
Notes. We quote the properties adopted from V09a and state the overdensity radii and corresponding cluster masses used for Figs. 1 and 2. All
masses are in units of 1014 M , without applying the E(z) factor. We state only stochastic uncertainties, i.e. do not include systematics. The proxyY,T,G
Y,T,G
based masses M500
from V09a in the first part of this table are defined in Sect. 2.5, as well as the r500
quoted in the second part. The third and
wl
hyd
fourth part of the table show the weak lensing (M ) and hydrostatic (M ) for different cases, respectively. By cfit and cB13 we denote the choices
for the NFW concentration parameter explained in Sect. 2.2. We refer to Sect. 4.3 for the introduction of the “no dilu. corr.” case (only differing
from the default for the first four clusters). See Table A.1 for further properties.
temperature profiles (barring only the two most exceptional
clusters). For r200 , we use the WL results from Paper II2 .
Equation (4) provides us with a cumulative mass profile.
We evaluate this profile at some rtest , e.g. from WL, and propagate the uncertainty in rtest , together with the uncertainty in T X .
Hydrostatic equilibrium and sphericity are known to be
problematic assumptions for many clusters. Nonetheless, hydrostatic masses are commonly used in the literature in comparisons
to WL masses. Our goal is to study if and how biases due to
deviations from the above-mentioned assumptions show up.
2.5. Mass estimates
Table 1 comprises the key results on radii r500 and the corresponding mass estimates. By M P (rQ ), we denote a mass measured from data on proxy P within a radius defined by proxy Q.
2
For the scaling relations within WL-derived radii, we choose the respective cNFW . Otherwise, we use cB13 as a default.
A129, page 4 of 15
We use five mass estimates: P, Q ∈ {wl, hyd, Y, T, G}. The first
two are the weak lensing (wl) and hydrostatic X-ray masses
(hyd), as introduced in Sects. 2.2 and 2.4. Having analysed deep
Chandra observations they acquired, Vikhlinin et al. (2009a)
present three further mass estimates for all 36 clusters in the
complete sample. Based on the proxies T X , the ICM mass Mgas ,
and YX = T X Mgas , mass estimates M T , M G , and M Y are quoted
in Table 2 of V09a. We point out that V09a obtain these estimates by calibrating the mass scaling relations for respective
proxy on local clusters (see their Table 3). V09a further provide a detailed account of the relevant systematic sources of
uncertainty.
1/3
P
P
The radii r500
= 3M500
/(2000πρc )
listed in Table 1 are
P
obtained from M500
, P ∈ {Y, T, G}. Using Eqs. (1) and (4), we
then derive the WL and hydrostatic masses, respectively, within
these radii. We emphasise that all WL mass uncertainties quoted
in Table 1 are purely statistical and do not include any of the
systematics discussed in Paper II.
H. Israel et al.: The 400d Galaxy Cluster Survey weak lensing programme. III.
2.6. Fitting algorithm for scaling relations
The problem of selecting the best linear representation y =
A + Bx for a sample of (astronomical) observations of two
quantities {xi } and {yi } can be surprisingly complex. A plethora
of algorithms and literature cope with the different assumptions about measurement uncertainties one can or has to make
(e.g. Press et al. 1992; Akritas & Bershady 1996; Tremaine
et al. 2002; Kelly 2007; Hogg et al. 2010; Williams et al. 2010;
Andreon & Hurn 2012; Feigelson & Babu 2012). The challenges observational astronomers have to tackle when trying to
reconcile the prerequisites of statistical estimators with the realities of astrophysical data are manifold, including heteroscedastic uncertainties (i.e. depending non-trivially on the data themselves), intrinsic scatter, poor knowledge of systematics, poor
sample statistics, “outlier” points, and non-Gaussian probability
distributions. Tailored to the problem of galaxy cluster scaling
relations, Maughan (2014) proposed a “self-consistent” modelling approach based on the fundamental observables. A full
account of these different effects exceeds the scope of this article. We choose the relatively simple fitexy algorithm (Press
et al. 1992), minimising the estimator
χ2P92 =
N
X
(yi − A − Bxi )2
i=1
σ2y,i + B2 σ2x,i
,
(5)
which allows the uncertainties σ x,i and σy,i to vary for different data points xi and yi , but assumes them to be drawn from a
Gaussian distribution. To accommodate intrinsic scatter, σ2y,i in
Eq. (5) can be replaced by σ2i = σ2y,i + σ2int (e.g. Weiner et al.
2006; Andreon & Hurn 2012). We test for intrinsic scatter using
mpfitexy (Markwardt 2009; Williams et al. 2010), but in most
cases, due to the small χ2 values, find the respective parameter
not invoked. Thus we decide against this additional complexity.
A strength of Eq. (5) is its invariance under changing x and y
(e.g. Tremaine et al. 2002); i.e., we do not assume either to be
“the independent variable”.
Rather than propagating the (unknown) distribution functions in the mass uncertainties3 , we approximate 1σ Gaussian
uncertainties in decadic log-space, applying the symmetrisation:
σ(log ξi ) = log (e)(ξi+ − ξi− )/(2ξi ) = log (e)(σ+ξ,i + σ−ξ,i )/(2ξi ),
(6)
where ξi+ = ξi + σ+ξ,i and ξi− = ξi − σ−ξ,i are the upper and
lower limits of the 1σ interval (in linear space) for the datum ξi ,
given the uncertainties σ±ξ,i . All our calculations and plots use
{xi } := {log ξi } and {σ x,i } := {σ(log ξi ) }, with log ≡ log10 .
3. Results
3.1. Weak lensing and hydrostatic masses
The first and single most important observation is that hydrohyd wl
static masses M500 (r500
), i.e. evaluated at r500 as found from
wl
wl
weak lensing, and weak lensing masses M500
(r500
) roughly agree
with each other (Table 1). Our second key observation is the
hyd
wl
very tight scaling behaviour between M500 and M500
, as Fig. 1
shows. In all cases presented in Fig. 1, and most of the ones
we tested, all data points are consistent with the best-fit relation. Consequently, the fits return small values of χ2red < 1 (see
3
A natural feature in complex measurements like this, asymmetric unwl
certainties in M200
arise from the grid approach to χ2 minimisation in
Paper II (cf. Fig. 2 therein).
wl
wl
Fig. 1. Scaling of weak lensing masses M500
(r500
) with hydrostatic
hyd wl
masses M500 (r500 ). The upper (lower) panel is for cfit (cB13 ). Both show
best fits for three cases: the default (filled, thick ring, dotted ring symbols; thick dashed line), regular shear profile clusters only (filled and
thick ring symbols; dash-dotted line; Sect. 4.1), and without correction
for dilution by cluster members (filled, thin ring, dotted ring symbols;
long dashed line; Sect. 4.3). The dotted line shows equality of the two
hyd
wl
masses, M500
= M500
. Shaded regions indicate the uncertainty range of
the default best-fit. Some error bars were omitted for sake of clarity.
Table 2). Bearing in mind that we only use stochastic uncertainties, this points to some intrinsic correlation of the WL and
hydrostatic masses. We will discuss this point in Sect. 4.2.
hyd wl
wl
wl
Finally, we find the slope of the M500
(r500
)–M500 (r500
) relation (dashed lines in Fig. 1) to be steeper than unity (dotted line):
using the “default model”, i.e. the analysis described in Sect. 2,
a fitexy fit yields 1.71 ± 0.64 for the “cfit ” case (concentration
A129, page 5 of 15
A&A 564, A129 (2014)
Table 2. Measurements of the X-ray – WL mass bias.
Scaling relation
hyd wl
wl
wl
M500
(r500
)–M500
(r500 )
M wl (rfix )–M hyd (rfix )
wl
Y
Y
Y
M500
(r500
)–M500
(r500
)
wl
T
T
T
M500 (r500 )–M500 (r500 )
wl
G
G
G
M500
(r500
)–M500
(r500
)
Model
default
default
no dilu. corr.
no dilu. corr.
cNFW
Intercept A
cfit
−0.51+0.20
0.00+0.07
−0.21
−0.08
+0.26
cB13 −0.47−0.25
−0.02+0.07
−0.08
cfit −0.53 ± 0.23 0.01 ± 0.08
cB13 −0.49 ± 0.29 −0.01+0.07
−0.08
rfix = 600 kpc cB13
rfix = 800 kpc cB13
rfix = 1000 kpc cB13
default
default
default
Slope B
−0.68+0.19
−0.21
−0.58+0.19
−0.21
−0.52+0.19
−0.21
bMC from Monte Carlo
0.08+0.14
−0.13
0.00+0.14
−0.13
+0.14
0.11−0.13
+0.14
0.02−0.13
b = hlog ξ − log ηi χ2red,M−M Section
+0.21
(0.27−0.20
; −0.10+0.16
−0.15 )
+0.23
(0.10−0.18 ; −0.10+0.17
−0.15 )
+0.21
(0.27−0.20
; −0.06+0.16
−0.15 )
+0.22
(0.10−0.18 ; −0.06+0.17
−0.15 )
0.08 ± 0.09
−0.02 ± 0.04
0.11 ± 0.08
0.00 ± 0.03
0.58
0.52
0.57
0.51
3.1
3.1
4.3
4.3
+0.10
+0.16
−0.11 ± 0.05 0.01−0.07
(0.12−0.10
; −0.11+0.10
−0.08 )
+0.10
+0.18
−0.02 ± 0.04 0.02−0.07 (0.12−0.11 ; −0.09+0.10
−0.08 )
+0.11
+0.11
0.01 ± 0.05 0.01−0.08
(0.10+0.20
;
−0.09
−0.11
−0.09 )
−0.02 ± 0.04
−0.02 ± 0.04
−0.03 ± 0.03
0.82
0.72
0.69
4.2
4.2
4.2
0.04 ± 0.06
0.02 ± 0.05
0.00 ± 0.07
1.21
0.88
2.11
3.2
3.2
3.2
+0.10
+0.10
cB13 −0.75+0.12
0.07 ± 0.03 0.08−0.07
(0.23+0.18
−0.13
−0.11 ; −0.08−0.07 )
+0.18
+0.11
cB13 −0.63 ± 0.23 0.04 ± 0.06 0.05+0.11
−0.08 (0.17−0.12 ; −0.08−0.10 )
+0.09
+0.18
+0.03
+0.10
+0.17
cB13 −0.89−0.31
0.01−0.04 0.04−0.07 (0.21−0.10 ; −0.15−0.07 )
Notes. We estimate a possible
bias between masses ξ and η by three estimators: First, we fit to (log ξ − log η) as a function of η, yielding
an intercept A at pivot log Mpiv /M = 14.5 and slope B from the Monte Carlo/jackknife analysis. Second, we compute the logarithmic bias
bMC = hlog ξ −log ηiMC , averaged over the same realisations. Uncertainties for the MC results are given by 1σ ensemble dispersions. In parentheses
next to bMC , we show its value for the low-M wl and high-M wl clusters. Third, we quote the logarithmic bias b = hlog ξ − log ηi obtained directly
from the input masses, along with its standard error. Finally, we give the χ2red for the mass-mass scaling, obtained from the MC method. The
“default” model denotes WL and hydrostatic masses as described in Sect. 2.
wl
Fig. 2. Ratios between X-ray and WL masses as a function of WL mass. Panel A) shows log (M hyd /M wl ) within r500
, panel B) shows log (M Y /M wl )
Y
within r500
. WL masses assume the B13 c–M relation. We show three tests for a mass bias: the overall average logarithmic bias b = hlog M X −
log M wl i is denoted by a long-dashed line, and its standard error by a dark grey shading. Short-dashed lines and light grey shading denote the
same quantity, but obtained from averaging over Monte Carlo realisations including the jackknife test. We also show this bMC for the low-M wl and
high-M wl clusters separately, with the 1σ uncertainties presented as boxes, for sake of clarity. As a visual aid, a dot-dashed line depicts the Monte
Carlo/jackknife best-fit of log (M X /M wl ) as a function of M wl . In addition, panel A) also contains this best-fit line (triple-dot-dashed) for the case
without correction for cluster member dilution; the corresponding data points follow the Fig. 1 scheme. Indicated by uncertainty bars, panel B)
also presents three high-z clusters from High et al. (2012).
parameters from the shear profile fits, cf. Paper II; upper panel of
Fig. 1), and 1.46±0.57, if the B13 mass-concentration relation is
applied (“cB13 ”; lower panel). The different slopes in the cfit and
cB13 cases are mainly due to the two clusters, CL 1641+4001
and CL 1701+6414, in which the weak lensing analysis revealed
shallow tangential shear profiles due to extended surface mass
plateaus (cf. Figs. 3 and 5 of Paper II). This will be the starting
point for further analysis and interpretation in Sect. 4.1.
A129, page 6 of 15
Although the cB13 slope is consistent with the expected
hyd
wl
1:1 relation, such a M500
–M500 relation would translate to extreme biases between X-ray and WL masses if extrapolated
to higher and lower masses. Especially for masses of a
few 1015 M , ample observations disagree with the extrapolated
M wl > 2M hyd . We do not claim our data to have such predicting
power outside its mass range. Rather, we focus on what can be
learnt about the X-ray/WL mass bias in our 0.4 ∼ z ∼ 0.5 mass
H. Israel et al.: The 400d Galaxy Cluster Survey weak lensing programme. III.
range, which we, for the first time, study in the mass range down
to ∼1 × 1014 M .
We are using three methods to test for the biases between
X-ray and WL masses. First, we compute the logarithmic bias
b = hlog ξ − log ηi, which we define as the average logarithmic difference between two general quantities ξ and η. Its
interpretation is that 10b η is the average value corresponding
to ξ. The uncertainty in b is given by the standard error of
(log ξ − log η). Hence, our measurement of b = −0.02 ± 0.04
for cB13 corresponds to a vanishing fractional bias of hM hyd i ≈
(0.97 ± 0.09)hM wl i.
Given the small sample size, large uncertainties, and the
tight scaling relations in Fig. 1 pointing to some correlation between the WL and X-ray masses, we base our further tests on a
Monte Carlo (MC) analysis including the jackknife test. For 105
realisations, we chose ξ̂i,k = ξi + δξi,k with random δξi,k drawn
from zero-mean distributions assembled from two Gaussian
halves with variances σ−ξ,i for the negative and σ+ξ,i for the positive half4 . This provides a simple way of accommodating asymmetric uncertainties (cf. Paper II and Table 1). Then we take
the logarithm and again symmetrise the errors. We repeat for
η̂i,k = ηi + δηi,k . On top, for each realisation {ξ̂i,k , η̂i,k }, we discard
one cluster after another, yielding a total of 8 × 105 samples.
Based on those MC/jackknife realisations, we compute our
second bias estimator bMC = hlog ξ̂ − log η̂iMC . In order to
achieve the best possible robustness against large uncertainties
and small cluster numbers, we quote the ensemble median and
dispersion. We find bMC = 0.00+0.14
−0.13 for cB13 , in good agreement
with b = −0.02 ± 0.04, i.e. a median WL/X-ray mass ratio of 1.
Fitting log (M X /M wl ) as a function of M wl and averaging
over the MC/jackknife samples, we obtain our
third bias
estimator, an intercept A at the pivot mass of log Mpiv /M = 14.5.
We find A = −0.02+0.07
−0.08 for cB13 , again consistent with vanishing
bias.
does not support this hypothesis. We point out that bMC was
designed to be both robust against possible effects of large
uncertainties and the small number of clusters in this first batch
of 400d WL clusters. The larger uncertainties in bMC compared
to b are directly caused by the jackknife test and the account for
possible fit instability in the MC method.
The slopes B quantifying the M wl dependence of the
X-ray/WL mass ratio are significantly negative in all of our measurements. This directly corresponds to the steep slope in the
mass-mass scaling (Fig. 1). Predicting cluster masses for very
massive clusters (or low-mass groups) from B = −0.75 for
M Y would yield M X = M wl /2 at 1015 M /E(z) (and M X =
2.8 M wl at 1014 M /E(z)). Such ratios are at odds with existing measurements. Therefore, we refrain from extrapolating
cluster masses, but interpret the slopes B as indicative of a
possibly mass-dependent X-ray/WL mass ratio. This evidence
is more prudently presented as the ∼2σ discrepancy between
the low-M wl and high-M wl mass bins for all three V09a X-ray
observables.
4. A mass-dependent bias?
In this section, we analyse two unexpected outcomes of our
study in greater detail: the clear correlation between the
M X /M wl measurement of the individual clusters and their lensing masses, and the unusually small scatter in our scaling relations. Results for ancillary scaling relations that we present in
Appendix A underpin the findings of Fig. 1 and Table 2. To begin
with, we emphasise that the mass-dependent bias is not caused
by the conflation of a large z range; all but one of our clusters
inhabit the range 0.39 ≤ z ≤ 0.53 across which E(z) varies by
<10%, and we accounted for this variation. As Fig. A.2 shows,
this also leaves us with little constraining power with regard to
a z-dependent bias, at least until the 400d WL survey becomes
more complete.
3.2. Lensing masses and X-ray masses from proxies
Figures 2 and A.2, as well as Table 2 present the three different X-ray/WL mass bias estimates for four X-ray mass obhyd
servables: M500 using cB13 from Sect. 3.1 in Panel A of Fig. 2,
Y
T
G
M500 in Panel B of Fig. 2, M500
in Panel A of Fig. A.2, M500
in
Panel B of Fig. A.2. The last three are the proxy-based Vikhlinin
et al. (2009a) X-ray mass estimators defined in Sect. 2.5. While
Y,T,G
wl
Panel A uses r500
, the other three panel use the respective r500
.
Long-dashed lines and dark-grey boxes in Fig. 2 display b and its error. Short-dashed lines and light-grey boxes
denote bMC . The intercept A is located at the intersection of
the dot-dashed fit and dotted zero lines. We also show bMC
for the low-mass and high-mass clusters separately, splitting
at M wl E(z) = Mpiv ; the respective 1σ ranges are shown as outline boxes.
We observe remarkable agreement between the
X-ray/WL mass ratios and bias fits from all four X-ray
observables (which are not fully independent). For each of
them, all three bias estimates agree with each other, and all are
consistent with no X-ray/WL mass bias. We find no evidence
for X-ray masses being biased low by ∼40% in our cluster
sample, as it has been suggested to explain the Planck CMB –
SZ cluster counts discrepancy. While bMC ≥ 0.2 (&35% mass
bias) lies within the possible range of the high-M wl bin, in
particular using the gas mass M G , the overall cluster sample
4
Unphysical cluster masses <1013 M /E(z) are set to 1013 M /E(z).
4.1. Role of c200 and departures from NFW profile
Figure 1 shows that the M wl –M X scaling relation sensitively depends on the choice for the cluster concentration parameter c200 .
This translates into more positive bias estimates for cfit as compared to cB13 (Table 2). The difference is caused by the two
flat-profile clusters for which NFW fits yield low masses but
do not capture all the large-scale mass distribution, in particular if c200 is determined directly from the data, rather than assuming a mass-concentration relation (cf. the discussion of the
concentration parameter in Paper II and Foëx et al. 2012). This
induces a bias towards low masses in the r200 → r500 conversion.
If these two “irregular” clusters (dotted ring symbols in Fig. 1)
are excluded, the “regular clusters only” M wl –M X scaling relations (dash-dotted lines) differ for the cfit , but not for the cB13
case. Moreover, their mass ratios are consistent with the other
high-M wl clusters for cB13 . We thus confirm that assuming a
mass-concentration relation and marginalising over c200 is advantageous for scaling relations. We note that Comerford et al.
(2010) observed a correlation between the scatters in the massconcentration and mass-temperature relations and advocated the
inclusion of unrelaxed clusters in scaling relation studies.
Because NFW profile fits do not capture the complete
projected mass morphologies of irregular clusters, the assumpwl
wl
wl
Y
), etc.
(r500
) (Eq. 3), and M500
(r500
tion of that profile for M500
(Sect. 3.2) could introduce a further bias. Aperture-based lensing masses, e.g. the ζc statistics (Clowe et al. 1998) employed by
A129, page 7 of 15
A&A 564, A129 (2014)
Hoekstra et al. (2012) provide an alternative. However, Okabe
et al. (2013) demonstrated by the stacking of 50 clusters from
the Local Cluster Substructure Survey (0.15 < z < 0.3), that the
average weak lensing profile does follow NFW to a high degree,
at least at low redshift. Furthermore, the Planck Collaboration
(2013) finds a trend of M wl /M hyd with the ratio of concentration
parameters measured from weak lensing and X-rays.
4.2. Correlation between mass estimates
In Table 2, we quote χ2red for the M wl –M X scaling relations, using
the same MC/jackknife samples as for the bias tests. For the ones
involving hydrostatic masses, we measure 0.5 < χ2red < 0.6. We
hyd
wl
evaluated M500 at r500
in order to measure both estimates within
the same physical radius, in an “apples with apples” comparison.
But using the lensing-derived radius also introduces an unknown
amount of correlation, a possible (partial) cause of the measured
low χ2red values.
We test for the impact of the correlation by measuring both
M wl and M hyd within a fixed physical radius for all clusters,
and choose rfix = 800 kpc as a rough sample average of r500 .
Surprisingly, we find an only slightly higher χ2red , still <1 (see
Table 2). As before, the bias estimators are consistent with zero.
Fixed radii of 600 kpc and 1000 kpc give similar results (Table 2
and Fig. A.2). with a tentative trend of increasing χ2red with
smaller radii. Interestingly, a low χ2red is also found for the M wl –
M T and M T –M hyd relations (Table A.2). The latter is expected,
because M hyd are derived from the same T X and depend sensitively on them. This all suggests that the small scatter is not
wl
driven by using r500
, but by some other intrinsic factor.
If the uncertainties in M wl were overestimated significantly,
this would obviously explain the low χ2red . However, we do not
even include systematic effects here. Moreover, the quoted M wl
uncertainties directly come from the NFW modelling of Paper II
and reflect the ∆χ2 from their Eq. (3), given the shear catalogue.
The errors are dominated by the intrinsic source ellipticity σε ,
for which we, after shear calibration, find values of ∼0.3, consistent with other ground-based WL experiments. Therefore, despite the allure of our WL errors being overestimated, we do
not find evidence for this hypothesis in our shear catalogues.
Furthermore, the quoted M wl uncertainties are consistent with
the aperture mass detection significances we reported in Paper II.
4.3. Dilution by cluster members and foregrounds
Comparing the complete set of mass-mass scaling relations our
data offer (Table A.2), we trace the mass-dependence of the bias
seen in Fig. 2 back to the different ranges spanned by the estimates for r500 . While the ratio of minimum to maximum is ≈0.75
hyd
Y
T
wl
for r500
, r500
, and r500 , the same ratio is 0.57 for r500
, using the
B13 M–c relation. In the following, we discuss the potential
influence of several sources of uncertainty in the WL masses,
wl
showing that the dispersion between lowest and highest M500
is
likely an inherent feature rather than a modelling artefact.
In Paper II, we discussed the great effort we took in constructing the best-possible homogeneous analysis from the quite
heterogeneous MMT imaging data. Unfortunately, we happen to
find higher M wl for all clusters with imaging in three bands than
for the clusters with imaging in one band. We emphasise there
are no trends with limiting magnitude, seeing, or density nKSB of
galaxies with measurable shape (cf. Tables 1 and 2 in Paper II).
A129, page 8 of 15
In the cases where three-band imaging is available, our
WL model includes a correction for the diluting effect residual
cluster member galaxies impose on the shear catalogue. For the
other clusters, no such dilution correction could be applied. A
rough estimate of the fraction of unlensed galaxies remaining
after background selection suggests that the contamination in
single-band catalogues is ∼30–50% higher than with the more
sophisticated galaxy-colour based method. Therefore, we recalculate the scaling relations, switching off the dilution correction (long dashed line and thin ring symbols in Fig. 1). This
lowers the r500 values by ∼6% and the masses by 10–15%. For
hyd
wl
wl
Y
both the M500
–M500 and M500
–M500
relations, we only observe
a slightly smaller difference between bMC for the high- and lowM wl bins (Tables 2 and A.2), not significant given the uncertainty
margins.
The dilution of the shear signal by an increased number
of galaxies not bearing a shear signal can also be expressed
as a overestimation of the mean lensing depth hβi. We model
a possible lensing depth bias by simultaneously adding the
uncertainty σ(hβi) for the three-band clusters and subtracting it
for the single-band ones, maximising the leveraging effect on
the masses. Similar to the previous experiment5 we still observe
a mass-dependent bias, with little change to the default model.
We further tested alternative choices of cluster centre and
fitting range, but do not observe significant changes to the mass
dependent-bias or to χ2red (see Appendix A.3). Although hβi is
calculated for all clusters from the same catalogues, related systematics would affect the mass normalisation, but not the relative
stochastic uncertainties, which determine χ2red . As we consider a
drastic overestimation of the purely statistic uncertainties in the
WL modelling being unlikely (Sect. 4.2), the cause of the low
χ2red values remains elusive. If a WL analysis effect is responsible for one or both anomalies, it has to be of a more subtle nature
than the choices investigated here.
4.4. A statistical fluke?
hyd
wl
We summarise that the M500
–M500 scaling relations we observe
show an unusual lack of scatter and that we find a ∼2σ difference
between the X-ray/WL mass ratios of our high- and low-M wl
clusters. The latter effect can be traced back to the considerable
span in cluster lensing signal, which is only partly due to different background selection procedures and the dilution correction
that was only applied for clusters imaged in three bands.
The question then arises if the mass-dependent bias
is caused by an unlucky selection of the 8 MMT clusters
from V09’s wider sample of 36. The 8 clusters were chosen to be
observed first merely because of convenient telescope scheduling, and appear typical of the larger sample in terms of redshift and X-ray observables. The MMT clusters trace well the
mass range and dispersion spanned by all 36 clusters in their
Y
T
M500
–M500
relation. We observe the expected vanishing slopes
for log (M T /M Y ) as a function of M Y , both for the 8 MMT clusters and for the complete sample of 36 (Table A.2).
In Table 2, we observe significant scatter (χ2red = 2.11) in
wl
G
the M500
–M500
relation, while Okabe et al. (2010b) and M 13 reported particularly low scatter in M G , comparing to WL masses.
This large intrinsic scatter seems to be a feature of the overall 400d sample: plotting M G versus the two other V09 X-ray
5
Because an unnoticed higher dilution in the catalogue does not imply
a bias in the estimation of hβi from a proxy catalogue, the two effects
are not likely to add up.
H. Israel et al.: The 400d Galaxy Cluster Survey weak lensing programme. III.
Fig. 3. Comparisons with literature data. Left panel: black symbols show z > 0.35 clusters from Mahdavi et al. (2013), whose best-fit using Eq. (5)
is shown by the dot-dashed line. The cluster CL 1524+0957 is indicated by a diamond symbol. Coloured symbols and the dashed line show the
hyd
wl
“default” M500
–M500
relation for cB13 as in the lower panel of Fig. 1. Middle panel: the same, but comparing to Foëx et al. (2012) (black symbols
hyd
and dot-dashed line for best fit). X-ray masses are measured within r500
. CL 1003+3253 and CL 1120+4318 are emphasised by special symbols.
wl
Right panel: scaling of lensing masses M500 with the YX proxy. Black symbols show the z > 0.35 clusters from M13, to which the thick, dash-dotted
line is the best fit. Shaded regions indicate the uncertainties to this fit. The thin, dash-dotted line gives the best fit M13 quote for their complete
sample, while the thin solid and long-dashed lines mark the M500 –YX relations by V09a and Arnaud et al. (2010), respectively, for z = 0.40.
masses of all 36 clusters, we also find χ2red > 2 (Table A.2),
as well as significant non-zero logarithmic biases. While tracing
the cause of this observation is beyond the scope of this article, it
deserves
Y,T further
study. Because two of the clusters with highest
G M500 − M500
are covered by our MMT subsample, we observe
a more mass-dependent M G /M Y,T ratio than for all 36. Overall,
however, the MMT subsample is not a very biased selection.
on different radial and mass scales, such that this explanation
cannot be ruled out. We summarise that the greater dynamical
range in WL than in X-ray masses might be linked to different
sensitivities of the respective methods to substructure and mergers in the low-mass, high-z cluster population we are probing,
but which is currently still underexplored.
5. Comparison with previous works
hyd
4.5. Physical causes
5.1. The Mwl
500 –M500 relation
An alternative and likely explanation for the mass-dependent
bias we observe could be a high rate of unrelaxed clusters, especially for our least massive objects. If the departure from
hydrostatic equilibrium were stronger among the low-mass
clusters than for the massive ones, this would manifest in
mass ratios similar to our results. Simulations show the offset from hydrostatic equilibrium to be mass-dependent (Rasia
et al. 2012), despite currently being focused on the high-mass
regime. Variability in the non-thermal pressure support with
mass (Laganá et al. 2013) may be exacerbated by small number statistics. At high z, the expected fraction of merging clusters, especially of major mergers, increases. Unrelaxed cluster
states are known to affect X-ray observables and, via the NFW
fitting, also lensing mass estimates. Indeed, the two most deviant
systems in Fig. 2 are CL 1416+4446 and the flat-profile “shear
plateau” cluster CL 1641+4001. Although the first shows an inconspicuous shear profile, we suspect it to be part of a possibly
interacting supercluster, based on the presence of two nearby
structures at the same redshift, detected in X-ray as well as in
our lensing maps (Paper II). Both these clusters are classified as
non-mergers in the recent Nurgaliev et al. (2013) study, introducing a new substructure estimator based on X-ray morphology.
However, WL and X-ray methods are sensitive to substructure
Comparison with Mahdavi et al. (2013) results: recently, M13
published scaling relations observed between the weak lensing
and X-ray masses for a sample of 50 massive clusters, partly
based on the brightest clusters from the Einstein Observatory
Extended Medium Sensitivity Survey (Gioia et al. 1990). Weak
lensing masses for the M13 sample have been obtained from
CFHT/Megacam imaging (Hoekstra et al. 2012), while the X-ray
analysis combines XMM-Newton and Chandra data. While the
median redshift is z = 0.23, the distribution extends to z = 0.55,
including 12 clusters at z > 0.35. Owing to their selection,
these 12 clusters lie above the 400d flux and luminosity cuts,
making them directly comparable to our sample.
hyd
wl
The left panel in Fig. 3 superimposes the M500
and M500 of
the M13 high-z clusters on our results. The two samples overlap
at the massive (&5 × 1014 M ) end, but the 400d objects probe
down to 1 × 1014 M for the first time at this z and for these scaling relations. The slopes of the scaling relations are consistent:
using Eq. (5), we measure BM−M = 1.13 ± 0.20 for the 12 M13
objects. A joint fit with the 400d clusters (BM−M = 1.46 ± 0.57)
yields BM−M = 1.15 ± 0.14 and a low χ2red = 0.54, driven by
our data. We note that the logarithmic bias of b = 0.10 ± 0.05
for the M13 high-z clusters corresponds to a (20 ± 10) % mass
bias, consistent with both the upper range of the 400d results and
A129, page 9 of 15
A&A 564, A129 (2014)
expectation from the literature (e.g. Laganá et al. 2010; Rasia
et al. 2012).
Calculating the Hogg et al. (2010, H10) likelihood which
Mahdavi et al. (2013) use, we find BH10 = 1.18+0.22
−0.20 and intrinsic scatter σint consistent with zero, confirming our above
results. If we, however, repeating our fits from Fig. 1 with the
H10 likelihood, we obtain discrepant results which highlight
the differences between the various regression algorithms (see
Sect. 2.6)6 .
CL 1524+0957 at z = 0.52 is the only cluster the 400d and
M13 samples share. Denoted by a black diamond in Fig. 3, its
masses from the M13 lensing and hydrostatic analyses blend in
wl
with the MMT 400d clusters. If it were included in the M500
–
hyd
M500 relation, it would not significantly alter the best fit, but we
caution that different analysis methods have been employed, e.g.
M13 reporting aperture lensing masses based on the ζc statistics.
Comparison with Jee et al. (2011) results: Jee et al. (2011,
J11) studied 14 very massive and distant clusters (0.83 < z <
hyd
wl
1.46) and found their WL and hydrostatic masses M200
and M200
to agree well, similar to our results. However, they caution that
hyd
their M200 were obtained by extrapolating a singular isothermal sphere profile. Because we doubt that the Chandra-based
Vikhlinin et al. (2006) model reliably describes the ICM out to
hyd
such large radii, we refrain from deriving M200 . Nevertheless,
we notice that our the J11 samples not only shows similar
wl
M200
than our most massive clusters, but also contains the only
two 400d clusters exceeding the redshift of CL 0230+1836,
CL 0152−1357 at z = 0.83 and CL 1226+3332 at z = 0.89. Their
planned re-analysis will improve the leverage of our samples at
the high-z end.
Comparison with Foëx et al. (2012) results: in the middle
panel of Fig. 3, we compare our results to 11 clusters from the
EXCPRES XMM-Newton sample, analysed by Foëx et al. (2012,
F12) and located at a similar redshift range (0.41 ≤ z ≤ 0.61)
as the bulk of our sample. Selected to be representative of the
cluster population at z ≈ 0.5, these objects have been studied
with XMM-Newton in X-rays and CFHT/Megacam in the optical. Foëx et al. (2012) explicitly quote hydrostatic and lensing masses within their respective radii; thus we also show the
hyd hyd
wl
wl
M500
(r500
)–M500 (r500 ). Again, the more massive of our clusters
resemble the F12 sources, with the 400d MMT sample extending towards lower masses. Indeed, F12 study two clusters which
are part of our sample: these, CL 1002+3253 at z = 0.42 and
CL 1120+4318 at z = 0.60 mark their lowest lensing mass obwl
jects. At similar M500
on either sides of the best-fit 400d scaling
relation, their inclusion with the quoted masses would have no
immediate effect on its slope, but slightly increase its scatter.
Bearing in mind that Fig. 3 (middle panel) compares quantities measured at different radii, we notice that the significantly
flat best fit regression line to the F12 cluster masses, showing
a larger dispersion in hydrostatic than in WL masses, as opposed to the 400d MMT clusters. The comparisons in Fig. 3 underscore that while being broadly compatible with each other,
different studies are shaped by the fine details of their sample
6
In fact, regression lines not only depend on the likelihood or definition of the best fit, but also on the algorithm used to find its extremum,
and, if applicable, how uncertainties are transferred from the linear to
the logarithmic domain. Thus, our H10 slopes agree with the ones the
web-tool provided by M13 yield, but produce different uncertainties.
A129, page 10 of 15
selection and analysis methods. We will conduct a more detailed comparison between our results and the ones of Foëx
et al. (2012) and Mahdavi et al. (2013) once we re-analysed
the CFHT/Megacam of the three overlapping clusters, having
already shown the MMT and CFHT Megacams to produce
consistent lensing catalogues (Paper II).
wl
Y
5.2. The M500
– M500
relation
The right panel of Fig. 3 investigates the scaling behaviour of
wl
M500
with YX by comparing the 400d MMT clusters to the
z > 0.35 clusters from M137 . The difference between the two
samples is more pronounced than in the left panel, with only
the low-mass end of the M13 sample, including CL 1524+0957,
overlapping with our clusters. None of the 400d MMT clusters deviates significantly from the M500 –YX relation applied
Y
by V09a in the derivation of the M500
masses we used. The
V09a M500 –YX relation based on Chandra data for low-z clusters
(Vikhlinin et al. 2006) is in close agreement to the M13 result for
their complete sample, as well as the widely used Arnaud et al.
(2010) M500 –YX relation. For the latter as well as V09a we show
the version with a slope fixed to the self-similar expectation of
B = 3/5. The best fit to the M13 z > 0.35 essentially yields the
same slope as the complete sample (B = 0.55 ± 0.09 compared
to BH10 = 0.56 ± 0.07, calculated with the H10 method). The
higher normalisation for the high-z subsample can be likely explained as Malmquist bias due to the effective higher mass limit
in the M13 sample selection. The incompatibility of the least
massive MMT clusters with this fit highlights that we sample
lower mass clusters, which, at the same redshift, are likely to
have different physical properties.
The YX proxy is the X-ray equivalent to the integrated pressure signal YSZ seen by SZ observatories. Observations confirm
a close YX –YSZ correlation, with measured departures from the
1:1 slope considered inconclusive (Andersson et al. 2011; Rozo
et al. 2012). Performing a cursory comparison with SZ observations, we included in Fig. 2A data for three z > 0.35 clusters
from High et al. (2012) (dashed uncertainty bars), taken from
their Fig. 6. The abscissa values for the High et al. (2012) clusters (SPT-CL J2022−6323, SPT-CL J2030−5638, and SPT-CL
J2135−5726) show masses based on YSZ , derived from South
Pole Telescope SZ observations (Reichardt et al. 2013). The M wl
are derived from observations with the same Megacam instrument we used for the 400d clusters, but after its transfer to the
Magellan Clay telescope at Las Campanas Observatory, Chile.
In good agreement with zero bias, the High et al. (2012) clusters
are consistent with some of the lower mass 400d clusters. This
result suggests that the YX –YSZ equivalence might hold once
larger samples at high z and low masses will become available.
6. Summary and conclusions
In this article, we analysed the scaling relation between WL and
X-ray masses for 8 galaxy clusters drawn from the 400d sample
of X-ray-luminous 0.35 ≤ z ≤ 0.89 clusters. WL masses were
measured from the Israel et al. (2012) MMT/Megacam data, and
X-ray masses were based on the V09a Chandra analysis. We
summarise our main results as follows:
7
Owing to the availability of data, we need to use different definitions
of r500 for the two data sets.
H. Israel et al.: The 400d Galaxy Cluster Survey weak lensing programme. III.
1. We probe the WL–X-ray mass scaling relation, in an unexplored region of the parameter space for the first time: the
z ∼ 0.4–0.5 redshift range, down to 1 × 1014 M .
2. Using several X-ray mass estimates, we find the WL and
X-ray masses to be consistent with each other. Most of our
clusters are compatible with the M X = M wl line.
3. Assuming the M wl not to be significantly biased, we do not
find evidence for a systematic underestimation of the X-ray
masses by ∼40%, as suggested as a possible solution to the
discrepancy between the Planck CMB constraints on Ωm
and σ8 (the normalisation of the matter power spectrum)
and the Planck SZ cluster counts (Planck Collaboration XX
2014). While our results favour a small WL-X-ray mass bias,
they are consistent with both vanishing bias and the ∼20%
favoured by studies of non-thermal pressure support.
4. For the mass-mass scaling relations involving M wl , we observe a surprisingly low scatter 0.5 < χ2red < 0.6, although we
use only stochastic uncertainties and allow for correlated errors via a Monte Carlo method. Because the errors in M wl are
largely determined by the intrinsic WL shape noise σε , we
however deem a drastic overestimation unlikely (Sect. 4.2).
For the scaling relations involving MG , however, we observe
a large scatter, contrary to Okabe et al. (2010b) and M13.
5. Looking in detail, there are intriguing indications for a massdependence of the WL-X-ray mass ratios of our relatively
low-mass z ∼ 0.4–0.5 clusters. We observe a mass bias in
the low–M wl mass bin at the ∼2σ level when splitting the
sample at log (Mpiv /M ) = 14.5 This holds for the masses
V09a report based on the YX , T X , and MG proxies.
derived from it (Sect. 5.1). The M13 and Foëx et al. (2012) samples include three 400d clusters with CFHT WL masses. These
clusters neither point to significantly higher scatter nor to a less
mass-dependent bias (Fig. 3). We are planning a re-analysis of
the CFHT data, having demonstrated in Paper II that lensing catalogues from MMT and CFHT are nicely compatible. Such reanalysis is going to be helpful to identify more subtle WL analysis effects potentially responsible for the steep slopes and tight
correlation of WL and X-ray masses.
An alternative explanation are intrinsic differences in the
low-mass cluster population. That the 400d MMT sample probes
to slightly lower masses (1 × 1014 M ) than M13 or F12 becomes especially obvious from the M wl –Y X relation (Fig. 3,
Sect. 5.2). Because the 400d sample is more representative of
the z ∼ 0.4–0.5 cluster population, it is likely to contain more
significant mergers relative to the cluster mass, skewing mass
estimates (Sect. 4.5). Hence, the 400d survey might be the first
to see the onset of a mass regime in which cluster physics and
substructure lead the WL-X-ray scaling to deviate from what
is known at higher masses. Remarkably, Giles et al. (2014) are
finding a different steep slopes in their low-mass WL-X-ray scaling analysis. Detailed investigations of how their environment
shapes clusters like CL 1416+4446 might be necessary to improve our understanding of the cluster population to be seen by
future cosmology surveys. Analysis systematics might also behave differently at lower masses. A turn for WL cluster science
towards lower mass objects, e.g. through the completion of the
400d WL sample, will help addressing the question of evolution
in lensing mass scaling relations.
The discrepant temperatures Chandra and XMM-Newton measure in clusters (Nevalainen et al. 2010; Schellenberger et al.
2012) could provide a possible avenue to reconcile Planck cluster properties with the Planck cosmology.
We thoroughly investigate possible causes for the massdependent bias and tight scaling relations. First (Sect. 4.1), we
confirm that by using a mass-concentration relation instead of
directly fitting c200 from WL, we already significantly reduced
the bias due to conversion from r200 to r500 . We emphasise that,
on average, the NFW shear profile represents a suitable fit for
the cluster population (cf. Okabe et al. 2013). Measuring M hyd
wl
within r500
induces correlation between the data points in Fig. 1.
Removing this correlation by plotting both masses within a fixed
physical radius, we still find small scatter (Sect. 4.2).
We notice that the mass range occupied by the M wl exceeds
the X-ray mass ranges. Partially, this higher WL mass range can
be explained by the correction for dilution by member galaxies, which could be applied only where colour information was
available (Paper II). Coincidentally, this is the case for the more
massive half of the MMT sample in terms of M wl , thus boosting
the range of measured WL masses (Sect. 4.3). This result underscores the importance of correcting for the unavoidable inhomogeneities in WL data due to the demanding nature of WL observations (cf. Applegate et al. 2014). We find no further indications
for biases via the WL analysis. Furthermore the tight scaling precludes strong redshift effects, and we find that our small MMT
subsample is largely representative of the complete sample of
36 clusters, judging from the M Y –M T relation (Sect. 4.4). For
the M Y –M G and M T –M G relations, significant scatter (χ2red > 2)
is present in the larger sample. The former relation also shows
indications for a significant bias of M Y ≈ 1.15M G .
Weak lensing and hydrostatic masses for the 400d MMT
clusters are in good agreement with the z > 0.35 part of
hyd
wl
the Mahdavi et al. (2013) sample and the M500
–M500 relation
Acknowledgements. The authors express their thanks to M. Arnaud and
EXCPRES collaboration (private communication) for providing the hydrostatic
masses of the Foëx et al. (2012) clusters. We further thank A. Mahdavi for
providing the masses of the Mahdavi et al. (2013) clusters via their helpful
online interface. H.I. likes to thank M. Klein, J. Stott, and Y.-Y. Zhang, and
the audiences of his presentations for useful comments. The authors thank the
anonymous referee for their constructive suggestions. H.I. acknowledges support for this work has come from the Deutsche Forschungsgemeinschaft (DFG)
through Transregional Collaborative Research Centre TR 33 as well as through
the Schwerpunkt Program 1177 and through European Research Council
grant MIRG-CT-208994. T.H.R. acknowledges support by the DFG through
Heisenberg grant RE 1462/5 and grant RE 1462/6. T.E. is supported by the DFG
through project ER 327/3-1 and by the Transregional Collaborative Research
Centre TR 33 “The Dark Universe”. R.M. is supported by a Royal Society
University Research Fellowship. We acknowledge the grant of MMT observation time (program 2007B-0046) through NOAO public access. MMT time was
also provided through support from the F. H. Levinson Fund of the Silicon Valley
Community Foundation.
Note added in proof. After this paper was accepted, another
paper (von der Linden et al. 2014b) appeared as submitted,
pointing at the combined effects of hydrostatic mass bias and
calibration systematics. The cross calibration effect on cosmological parameter constraints is currently being tested directly
by Schellenberger et al. (in prep.). Rozo et al. (2014a,b) review and cross-calibrate the various scaling relations involved.
Alternatively, Burenin (2013) suggested additional massive neutrino species.
References
Akritas, M. G., & Bershady, M. A. 1996, ApJ, 470, 706
Allen, S. W., Evrard, A. E., & Mantz, A. B. 2011, ARA&A, 49, 409
Amendola, L., Appleby, S., Bacon, D., et al. 2013, Liv. Rev. Rel., 16, 6
Andersson, K., Benson, B. A., Ade, P. A. R., et al. 2011, ApJ, 738, 48
Andreon, S., & Hurn, M. A. 2012 [arXiv:1210.6232]
Applegate, D. E., von der Linden, A., Kelly, P. L., et al. 2014, MNRAS, 439, 48
Arnaud, M., Pratt, G. W., Piffaretti, R., et al. 2010, A&A, 517, A92
A129, page 11 of 15
A&A 564, A129 (2014)
Bahé, Y. M., McCarthy, I. G., & King, L. J. 2012, MNRAS, 421, 1073
Bartelmann, M. 1996, A&A, 313, 697
Becker, M. R., & Kravtsov, A. V. 2011, ApJ, 740, 25
Bharadwaj, V., Reiprich, T. H., Schellenberger, G., et al. 2014, A&A, accepted
[arXiv:1402.0868]
Bhattacharya, S., Habib, S., Heitmann, K., & Vikhlinin, A. 2013, ApJ, 766, 32
(B13)
Bullock, J. S., Kolatt, T. S., Sigad, Y., et al. 2001, MNRAS, 321, 559 (B01)
Burenin, R. A. 2013, Astron. Lett., 39, 357
Burenin, R. A., Vikhlinin, A., Hornstrup, A., et al. 2007, ApJS, 172, 561
Cavaliere, A., & Fusco-Femiano, R. 1978, A&A, 70, 677
Clowe, D., Luppino, G. A., Kaiser, N., Henry, J. P., & Gioia, I. M. 1998, ApJ,
497, L61
Comerford, J. M., Moustakas, L. A., & Natarajan, P. 2010, ApJ, 715, 162
Corless, V. L., & King, L. J. 2009, MNRAS, 396, 315
Croston, J. H., Pratt, G. W., Böhringer, H., et al. 2008, A&A, 487, 431
Dietrich, J. P., Erben, T., Lamer, G., et al. 2007, A&A, 470, 821
Dietrich, J. P., Böhnert, A., Lombardi, M., Hilbert, S., & Hartlap, J. 2012,
MNRAS, 419, 3547
Eckmiller, H. J., Hudson, D. S., & Reiprich, T. H. 2011, A&A, 535, A105
Erben, T., van Waerbeke, L., Bertin, E., Mellier, Y., & Schneider, P. 2001, A&A,
366, 717
Erben, T., Schirmer, M., Dietrich, J. P., et al. 2005, Astron. Nachr., 326, 432
Ettori, S. 2013, Astron. Nachr., 334, 354
Ettori, S., Donnarumma, A., Pointecouteau, E., et al. 2013, Space Sci. Rev., 177,
119
Feigelson, E., & Babu, G. 2012, Modern Statistical Methods for Astronomy:
With R Applications (Cambridge University Press)
Foëx, G., Soucail, G., Pointecouteau, E., et al. 2012, A&A, 546, A106 (F12)
George, M. R., Leauthaud, A., Bundy, K., et al. 2012, ApJ, 757, 2
Giles, P. A., Maughan, B. J., Hamana, T., et al. 2014 [arXiv:1402.4484]
Giodini, S., Lovisari, L., Pointecouteau, E., et al. 2013, Space Sci. Rev., 177, 247
Gioia, I. M., Maccacaro, T., Schild, R. E., et al. 1990, ApJS, 72, 567
Hartlap, J., Schrabback, T., Simon, P., & Schneider, P. 2009, A&A, 504, 689
Heymans, C., van Waerbeke, L., Bacon, D., et al. 2006, MNRAS, 368, 1323
High, F. W., Hoekstra, H., Leethochawalit, N., et al. 2012, ApJ, 758, 68
Hildebrandt, H., Erben, T., Dietrich, J. P., et al. 2006, A&A, 452, 1121
Hoekstra, H. 2007, MNRAS, 379, 317
Hoekstra, H., Donahue, M., Conselice, C. J., McNamara, B. R., & Voit, G. M.
2011a, ApJ, 726, 48
Hoekstra, H., Hartlap, J., Hilbert, S., & van Uitert, E. 2011b, MNRAS, 412, 2095
Hoekstra, H., Mahdavi, A., Babul, A., & Bildfell, C. 2012, MNRAS, 427, 1298
Hoekstra, H., Bartelmann, M., Dahle, H., et al. 2013, Space Sci. Rev., 177, 75
Hogg, D. W., Bovy, J., & Lang, D. 2010, unpublished
[arXiv:1008.4686] (H10)
Ilbert, O., Arnouts, S., McCracken, H. J., et al. 2006, A&A, 457, 841
Israel, H., Erben, T., Reiprich, T. H., et al. 2010, A&A, 520, A58 (Paper I)
Israel, H., Erben, T., Reiprich, T. H., et al. 2012, A&A, 546, A79 (Paper II)
Jee, M. J., Dawson, K. S., Hoekstra, H., et al. 2011, ApJ, 737, 59 (J11)
Kaiser, N. 1986, MNRAS, 222, 323
Kaiser, N., Squires, G., & Broadhurst, T. 1995, ApJ, 449, 460
Kelly, B. C. 2007, ApJ, 665, 1489
Laganá, T. F., de Souza, R. S., & Keller, G. R. 2010, A&A, 510, A76
Laganá, T. F., Martinet, N., Durret, F., et al. 2013, A&A, 555, A66
Laureijs, R., Amiaux, J., Arduini, S., et al. 2011
[arXiv:1110.3193]
Limousin, M., Morandi, A., Sereno, M., et al. 2013, Space Sci. Rev., 177, 155
Mahdavi, A., Hoekstra, H., Babul, A., & Henry, J. P. 2008, MNRAS, 384, 1567
Mahdavi, A., Hoekstra, H., Babul, A., et al. 2013, ApJ, 767, 116, (M13)
Markwardt, C. B. 2009, in Astronomical Data Analysis Software and
Systems XVIII, eds. D. A. Bohlender, D. Durand, & P. Dowler, ASP Conf.
Ser., 411, 251
Marrone, D. P., Smith, G. P., Richard, J., et al. 2009, ApJ, 701, L114
Marrone, D. P., Smith, G. P., Okabe, N., et al. 2012, ApJ, 754, 119
Massey, R., Hoekstra, H., Kitching, T., et al. 2013, MNRAS, 429, 661
Maughan, B. J. 2014, MNRAS, 437, 1171
Maughan, B. J., Giles, P. A., Randall, S. W., Jones, C., & Forman, W. R. 2012,
MNRAS, 421, 1583
Merloni, A., Predehl, P., Becker, W., et al. 2012
[arXiv:1209.3114]
Miyatake, H., Nishizawa, A. J., Takada, M., et al. 2013, MNRAS, 429, 3627
Nandra, K., Barret, D., Barcons, X., et al. 2013
[arXiv:1306.2307]
Navarro, J. F., Frenk, C. S., & White, S. D. M. 1996, ApJ, 462, 563
Navarro, J. F., Frenk, C. S., & White, S. D. M. 1997, ApJ, 490, 493
Nevalainen, J., David, L., & Guainazzi, M. 2010, A&A, 523, A22
Newman, A. B., Treu, T., Ellis, R. S., et al. 2013, ApJ, 765, 24
Nurgaliev, D., McDonald, M., Benson, B. A., Stubbs, C. W., & Vikhlinin, A.
2013, ApJ, 779, 112
Oguri, M., & Hamana, T. 2011, MNRAS, 414, 1851
Okabe, N., Takada, M., Umetsu, K., Futamase, T., & Smith, G. P. 2010a, PASJ,
62, 811
Okabe, N., Zhang, Y.-Y., Finoguenov, A., et al. 2010b, ApJ, 721, 875
Okabe, N., Smith, G. P., Umetsu, K., Takada, M., & Futamase, T. 2013, ApJ,
769, L35
Pacaud, F., Pierre, M., Adami, C., et al. 2007, MNRAS, 382, 1289
Pillepich, A., Porciani, C., & Reiprich, T. H. 2012, MNRAS, 422, 44
Planck Collaboration 2013, A&A, 550, A129
Planck Collaboration XX. 2014, A&A, in press
DOI: 10.1051/0004-6361/201321521
Pointecouteau, E., Reiprich, T. H., Adami, C., et al. 2013
[arXiv:1306.2319]
Postman, M., Coe, D., Benítez, N., et al. 2012, ApJS, 199, 25
Predehl, P., Andritschke, R., Böhringer, H., et al. 2010, in SPIE Conf. Ser., 7732
Press, W. H., Teukolsky, S. A., Vetterling, W. T., & Flannery, B. P. 1992,
Numerical recipes in C. The art of scientific computing, 2nd ed. (Cambridge:
University Press)
Rasia, E., Meneghetti, M., Martino, R., et al. 2012, New J. Phys., 14, 055018
Reese, E. D., Mroczkowski, T., Menanteau, F., et al. 2012, ApJ, 751, 12
Reichardt, C. L., Stalder, B., Bleem, L. E., et al. 2013, ApJ, 763, 127
Reichert, A., Böhringer, H., Fassbender, R., & Mühlegger, M. 2011, A&A, 535,
A4
Reiprich, T. H., Basu, K., Ettori, S., et al. 2013, Space Sci. Rev., 195
Rozo, E., Vikhlinin, A., & More, S. 2012, ApJ, 760, 67
Rozo, E., Bartlett, J. G., Evrard, A. E., & Rykoff, E. S. 2014a, MNRAS, 438, 78
Rozo, E., Rykoff, E. S., Bartlett, J. G., & Evrard, A. 2014b, MNRAS, 438, 49
Sarazin, C. L. 1988, X-ray emission from clusters of galaxies (Cambridge
University Press)
Schellenberger, G., Reiprich, T., & Lovisari, L. 2012, in Galaxy Clusters as Giant
Cosmic Laboratories, ed. J.-U. Ness, 54
Schirmer, M. 2013, ApJS, 209, 21
Schneider, P., King, L., & Erben, T. 2000, A&A, 353, 41
Schrabback, T., Erben, T., Simon, P., et al. 2007, A&A, 468, 823
Shi, X., & Komatsu, E. 2014 [arXiv:1401.7657]
Short, C. J., Thomas, P. A., Young, O. E., et al. 2010, MNRAS, 408,
2213
Stanek, R., Rasia, E., Evrard, A. E., Pearce, F., & Gazzola, L. 2010, ApJ, 715,
1508
Stott, J. P., Hickox, R. C., Edge, A. C., et al. 2012, MNRAS, 422, 2213
Tremaine, S., Gebhardt, K., Bender, R., et al. 2002, ApJ, 574, 740
Vikhlinin, A., Kravtsov, A., Forman, W., et al. 2006, ApJ, 640, 691
Vikhlinin, A., Burenin, R. A., Ebeling, H., et al. 2009a, ApJ, 692, 1033 (V09a)
Vikhlinin, A., Kravtsov, A. V., Burenin, R. A., et al. 2009b, ApJ, 692, 1060
von der Linden, A., Allen, M. T., Applegate, D. E., et al. 2014a, MNRAS, 439,
2
von der Linden, A., Mantz, A., Allen, S. W., et al. 2014b [arXiv:1402.2670]
Weiner, B. J., Willmer, C. N. A., Faber, S. M., et al. 2006, ApJ, 653, 1049
Williams, M. J., Bureau, M., & Cappellari, M. 2010, MNRAS, 409, 1330
Woody, D., Padin, S., Chauvin, E., et al. 2012, in SPIE Conf. Ser., 8444
Wright, C. O., & Brainerd, T. G. 2000, ApJ, 534, 34
Zhang, Y.-Y., Finoguenov, A., Böhringer, H., et al. 2008, A&A, 482, 451
Zhang, Y.-Y., Okabe, N., Finoguenov, A., et al. 2010, ApJ, 711, 1033
Pages 13 to 15 are available in the electronic edition of the journal at http://www.aanda.org
A129, page 12 of 15
H. Israel et al.: The 400d Galaxy Cluster Survey weak lensing programme. III.
relation, which they include in the model. We emphasise that the
Malmquist bias correction – which is not included here – applied
by V09a moves the clusters upwards in Fig. A.1, such that the
sample agrees with the best-fit from the low-z sample, as Fig. 12
in V09a demonstrates.
As already seen in Fig. 2, the M wl (large symbols in
Fig. A.1) and M Y agree well. Thus we can conclude that the
WL masses are consistent with the expectations from their LX .
Finally, we remark that the higher X-ray luminosities for the
some of the same clusters reported by Maughan et al. (2012)
in their study of the LX –T X relation are not in disagreement with
V09a, as Maughan et al. (2012) used bolometric luminosities.
A.2. Redshift scaling and cross-scaling of X-ray masses
Fig. A.1. Lensing mass – X-ray luminosity relation. The M–LX relation
wl
wl
Y
Y
is shown, for both M500
(r500
) (filled circles) and M500
(r500
) (small triangles). Open triangles represent the sample clusters for which MMT
lensing masses are not available. The V09a M–LX relation at z = 0.40
(z = 0.80) is denoted by a long-dashed black (short-dashed red) line.
Shaded (hatched) areas show the respective 1σ intrinsic scatter ranges.
Appendix A: Further scaling relations and tests
A.1. The LX –M relation
To better assess the consistency of our weak lensing masses with
the Vikhlinin et al. (2009a) results, we compare them to the LX –
Y
masses of their
MY -relation derived by V09a using the M500
low-z cluster sample. Figure A.1 inverts this relation by showing
wl
wl
the M500
(r500
) masses as a function of the 0.5–2.0 keV Chandra
luminosities measured by V09a. Statistical uncertainties in the
Chandra fluxes and, hence, luminosities are negligible for our
purposes. We calculate the expected 68 % confidence ranges in
mass for a given luminosity by inverting the scatter in LX at a
fixed M Y as given in Eq. (22) of V09a. For two fiducial redshifts, z = 0.40 and z = 0.80, spanning the unevenly populated
redshift range of the eight clusters, the M–LX relations and their
expected scatter are shown in Fig. A.1. Small filled triangles in
Y
Fig. A.1 show the M500
masses from which V09a derived the
LX –M relation. Our 8 MMT clusters are nicely tracing the distribution of the overall sample of 36 clusters (open triangles).
As an important step in the calculation of the mass function,
these authors show that their procedure is able to correct for the
Malmquist bias even in the presence of evolution in the LX –M
Here we show further results mentioned in the main body of the
article. Figure A.2 shows two examples of the X-ray/WL mass
ratio as a function of redshift. Owing to the inhomegenous redshift coverage of our clusters, we cannot constrain a redshift evolution. All of our bias estimates are consistent with zero bias.
Table A.2 shows the fit results and bias estimates for various tests we performed modifying our default model, as well as
for ancillary scaling relations. In particular, we probe the scaling behaviour of hydrostatic masses against the V09a estimates,
for which we find a M Y /M hyd tentatively biased high by ∼15%,
while M T and M G do not show similar biases.
A.3. Choice of centre and fitting range
Weak lensing masses obtained from profile fitting have been
shown to be sensitive to the choice of the fitting range (Becker &
Kravtsov 2011; Hoekstra et al. 2011b; Oguri & Hamana 2011).
Taking these results into account, we fitted the WL masses
within a fixed physical mass range. Varying the fitting range by
using rmin = 0 instead of 0.2 Mpc in one and rmax = 4.0 Mpc instead of 5.0 Mpc in another test, we find no evidence for a crucial
influence on our results.
Both simulations and observations establish (e.g. Dietrich
et al. 2012; George et al. 2012) that WL masses using lensing
cluster centres are biased high due to random noise with respect
to those based on independently obtained cluster centres, e.g.
hyd
wl
the ROSAT centres we employ. The fact that the M500
–M500 relation gives slightly milder difference between bMC for the highand low-M wl bins when the peak of the S -statistics is assumed as
the cluster centre (Table A.2) can be explained by the larger relative M wl “boost” for clusters with larger offset between X-ray
and lensing peaks. This affects the flat-profile clusters (Sect. 4.1)
in particular, translating into a greater effect for the cfit case than
for cB13 -based masses. We find that WL cluster centres only
slightly alleviate the observed mass-dependence.
A129, page 13 of 15
A&A 564, A129 (2014)
T
G
Fig. A.2. Continuation of Fig. 2. Panel A) shows log (M T /M wl ) within r500
, panel B) shows log (M G /M wl ) within r500
. Like panel A) of Fig. 2,
hyd
wl
panel C) presents log (M /M ), but showing both WL masses measured at a fixed physical radius rfix . Filled dots and dot-dashed lines correspond
to rfix = 800 kpc, while triangles and triple-dot-dashed lines denote rfix = 600 kpc. Uncertainties for the 600 kpc case were omitted for clarity.
Panel D) shows log (M hyd /M wl ) from Fig. 2 as a function of redshift. Thin solid lines indicating the 1σ uncertainty range of the best-fit Monte
Carlo/jackknife regression line (dot-dashed).
A129, page 14 of 15
H. Israel et al.: The 400d Galaxy Cluster Survey weak lensing programme. III.
Table A.1. Continuation of Table 1.
CL 0030
CL 0159
CL 0230
CL 0809
CL 1357
CL 1416
CL 1641
CL 1701
+2618
+0030
+1836
+2811
+6232
+4446
+4001
+6414
895+84
−90
829+142
−175
909+159
−191
1075+122
−135
710+136
−161
650+104
−123
687+136
−181
668+117
−143
3.70+1.14
−1.01
3.57+0.73
−0.76
3.71+0.87
−0.76
3.28+0.72
−0.65
3.41+0.66
−0.73
3.32+1.07
−0.96
2.58+1.57
−1.32
2.61+0.91
−1.09
3.78+1.27
−1.55
3.43+1.13
−1.33
2.48+0.86
−1.05
2.44+1.11
−0.99
5.54+3.45
−2.81
5.33+1.89
−2.15
5.30+2.55
−2.15
4.71+2.21
−1.89
4.67+1.68
−1.87
4.54+2.57
−2.12
5.70+2.17
−1.89
4.37+1.10
−1.13
4.08+1.91
−1.21
3.65+1.53
−1.04
4.70+1.12
−1.17
3.86+1.31
−1.17
1.91+1.32
−1.03
1.99+0.79
−0.94
2.08+0.84
−0.87
1.26+0.71
−0.59
1.46+0.51
−0.57
1.52+0.56
−0.63
1.60+1.15
−0.96
1.66+0.73
−0.97
1.05+0.35
−0.27
1.45+0.90
−0.75
–
–
–
–
wl
r500
(cB13 ), varying hβi [kpc]
wl
wl
M500
(r500
), using cB13 ,
M wl (rfix = 800 kpc)
varying hβi
wl
Y
M500
(r500
), using cfit
wl
Y
M500 (r500
), using cfit , no dilu. corr.
wl
Y
M500 (r500 ), using cB13 , no dilu. corr.
hyd wl
M500
(r500 ), using cB13 , varying hβi
hyd
M (rfix = 800 kpc)
hyd Y
M500
(r500 )
hyd T
M500 (r500 )
hyd G
M500
(r500 )
1.65+0.59
−0.69
1.42+0.55
−0.49
–
–
–
–
2.30+0.96
−0.88
1.31+0.52
−0.45
1.70+0.73
−0.73
1.82+0.52
−0.56
3.06 ± 0.61 2.39 ± 0.54 3.84 ± 0.71 2.89 ± 0.51 2.51 ± 0.38 1.62 ± 0.23 1.88 ± 0.35 2.01 ± 0.21
3.29+0.97
−0.88
2.44+0.85
−0.77
3.67+1.34
−1.14
3.41+0.94
−0.86
2.55+0.60
−0.58
1.66+0.45
−0.42
1.73+0.56
−0.54
2.15+0.38
−0.37
3.52+0.95
−0.88
2.48+0.84
−0.76
3.73+1.34
−1.14
3.16+0.98
−0.89
2.52+0.61
−0.59
1.47+0.52
−0.46
1.74+0.56
−0.54
2.04+0.41
−0.40
2.83+1.05
−0.93
2.26+0.90
−0.79
3.21+1.38
−1.13
3.49+0.93
−0.86
2.42+0.64
−0.62
1.77+0.42
−0.40
1.64+0.61
−0.60
2.13+0.39
−0.37
Notes. Further measured properties of the 400d MMT cluster sample. All masses are in units of 1014 M , without applying the E(z) factor. We
state only stochastic uncertainties, i.e. do not include systematics. By cfit and cB13 we denote the choices for the NFW concentration parameter
explained in Sect. 2.2. We refer to Sect. 4.3 for the introduction of the “varying hβi” case.
Table A.2. Continuation of Table 2.
Scaling relation
hyd wl
wl
wl
M500
(r500
)–M500
(r500 )
Model
cNFW Slope B
varying hβi
cfit −0.54+0.22
−0.21
varying hβi
cB13
S -peak centred
cfit
−0.50+0.30
−0.30
−0.48+0.24
−0.23
−0.47+0.26
−0.25
S -peak centred cB13
wl
Y
Y
Y
M500
(r500
)–M500
(r500
)
hyd Y
wl
Y
M500
(r500
)–M500
(r500 )
hyd Y
wl
Y
(r500 )
M500
(r500
)–M500
hyd T
wl
T
(r500 )
M500
(r500
)–M500
hyd G
wl
G
M500 (r500 )–M500 (r500 )
hyd Y
Y
Y
M500
(r500 )–M500
(r500
)
hyd T
T
T
)
M500 (r500 )–M500 (r500
hyd G
G
G
M500 (r500 )–M500 (r500 )
Y
Y
T
T
M500
(r500
)–M500
(r500
)
default
cfit
no dilu. corr.
cfit
no dilu. corr.
cB13
default
cB13
default
cB13
default
cB13
default
–
default
–
–
default (MMT8)
–
all 36
–
Y
Y
G
G
M500
(r500
)–M500
(r500
) default (MMT8)
all 36
T
T
G
G
M500
(r500
)–M500
(r500
) default (MMT8)
all 36
bMC from Monte Carlo
0.00+0.07
−0.08
+0.20
+0.16
0.08+0.14
−0.13 (0.23−0.18 ; −0.07−0.15 )
0.07 ± 0.07
−0.02 ± 0.08
+0.22
+0.17
0.01+0.15
−0.13 (0.08−0.18 ; −0.07−0.15 )
+0.19
+0.15
−0.03+0.13
−0.11 (0.08−0.16 ; −0.14−0.14 )
+0.12
+0.16
−0.05−0.10 (0.01−0.14 ; −0.12+0.15
−0.14 )
−0.01 ± 0.03
0.50
4.3
−0.04 ± 0.04
0.59
A.3
−0.07 ± 0.03
0.60
A.3
0.06 ± 0.07
1.67
3.2
0.09 ± 0.06
1.51
4.3
0.06 ± 0.05
1.08
4.3
−0.02 ± 0.07
−0.03 ± 0.07
0.05 ± 0.04
0.06 ± 0.04
0.07 ± 0.03
–
–
–
–
−0.56+0.23
−0.23
−0.53+0.23
−0.22
−0.61+0.24
−0.25
−0.76+0.21
−0.29
−0.53+0.37
−0.39
−1.20+0.36
−0.62
−0.07+0.65
−0.51
−0.05+0.13
−0.12
−0.39+0.47
−0.55
−0.05+0.06
−0.06
−1.24+0.28
−0.41
−0.55+0.10
−0.14
0.08+0.08
−0.06
0.11+0.08
−0.06
+0.10
0.10−0.07
+0.11
(0.26−0.08
;
+0.11
(0.26−0.08 ;
(0.23+0.18
−0.11 ;
−0.10+0.10
−0.08 )
−0.05+0.10
−0.08 )
−0.04+0.10
−0.07 )
0.60
4.3
+0.12
+0.12
0.02+0.09
−0.08 (0.14−0.10 ; −0.12−0.11 )
0.01 ± 0.05
0.77
3.2
0.00 ± 0.05
+0.18
+0.12
0.02+0.11
−0.09 (0.12−0.12 ; −0.10−0.10 )
−0.02 ± 0.04
0.65
3.2
0.04 ± 0.05
+0.09
+0.07
0.03+0.06
−0.05 (0.09−0.06 ; −0.05−0.08 )
−0.02 ± 0.04
0.66
3.2
0.01+0.11
−0.09
−0.02 ± 0.04
0.62
3.2
cfit −0.63+0.20
−0.20 0.01 ± 0.05
default
default
−0.73+0.13
−0.14
−0.71+0.14
−0.15
−0.73+0.13
−0.15
b = hlog ξ − log ηi χ2red,M−M Section
Intercept A
−0.01 ± 0.06
(0.12+0.18
−0.12 ;
−0.10+0.12
−0.12 )
0.07 ± 0.03
0.07 ± 0.05
0.06 ± 0.03
1.29
3.2
0.04 ± 0.05
0.04 ± 0.07
0.04 ± 0.02
0.79
3.2
−0.05+0.03
−0.04
0.03+0.06
−0.05
0.01 ± 0.05
1.85
3.2
−0.03+0.06
−0.07
−0.04 ± 0.05
−0.03 ± 0.03
1.27
4.4
0.00 ± 0.02
−0.03 ± 0.02
−0.02 ± 0.01
1.32
4.4
−0.02+0.05
−0.04
−0.07 ± 0.02
−0.07 ± 0.04
3.43
4.4
−0.05 ± 0.01
−0.07 ± 0.01
−0.07 ± 0.01
2.32
4.4
0.05 ± 0.03
−0.03+0.05
−0.04
−0.04 ± 0.06
2.81
4.4
0.02 ± 0.02
−0.04 ± 0.02
−0.05 ± 0.02
2.58
4.4
Notes. We estimate a possible
bias between masses ξ and η by three estimators: First, we fit to (log ξ − log η) as a function of η, yielding
an intercept A at pivot log Mpiv /M = 14.5 and slope B from the Monte Carlo/jackknife analysis. Second, we compute the logarithmic bias
bMC = hlog ξ −log ηiMC , averaged over the same realisations. Uncertainties for the MC results are given by 1σ ensemble dispersions. In parentheses
next to bMC , we show its value for the low-M wl and high-M wl clusters. Third, we quote the logarithmic bias b = hlog ξ − log ηi obtained directly
from the input masses, along with its standard error. Finally, we give the χ2red for the mass-mass scaling, obtained from the MC method. The
“default” model denotes WL and hydrostatic masses as described in Sect. 2.
A129, page 15 of 15